In this paper, the EKPCA algorithm with a better clustering effect is obtained by improving the metric distance formula of the K-prototypes algorithm, which is combined with the XGBoost algorithm with integration ability to construct a method for analyzing students’ ideological and political behavior. After verifying the validity of the method, the method is used to analyze students’ comprehensive quality, academic performance mental health status, etc., to help the innovation of ideological education. Students’ comprehensive quality assessment analysis scores on average fall within the range of 67-68 points, which is below average. There is a 3.5-point difference between the grades of 2022 freshman students and those of 2021 freshman students, and the average number of failed courses is twice as low as before. The decisiveness and cowardice score of male students is 5.28 points higher than that of female students which is 1.31 points, and the score of male students in terms of extroversion and introversion is 2.85 points higher than that of female students, male students are more extroverted and bold than female students, and more attention should be paid to female students. We aim to actively strengthen the guiding position of Marxist theory in Civic and Political Education, while also promoting the integration and innovative development of big data in Civic and Political Education.